Does larger scale enhance carbon efficiency? Assessing the impact of corporate size on manufacturing carbon emission efficiency

Global energy-related CO2 emissions continue to rise, with China—the world’s largest emitter—facing particular challenges from its rapidly expanding manufacturing sector, which accounts for a disproportionate share of national emissions relative to GDP. Improving manufacturing carbon emission efficiency (MCEE) is critical to reconciling economic growth with environmental protection. The paper focuses on how enterprise size affects MCEE in China’s manufacturing sector, whether this effect is linear or nonlinear, and whether environmental regulation, trade openness, and R&D investment generate threshold effects in this relationship. Using sectoral data for 29 manufacturing subsectors (2012–2021), the study motivates the analysis by noting that larger firms may both emit more and yet achieve higher efficiency via better technology and management. The work aims to inform policy on leveraging firm scale, regulation, innovation, and openness to promote low-carbon, efficiency-driven development aligned with China’s carbon peaking and neutrality commitments.

The review covers two strands. First, measurement of MCEE: single-factor approaches (e.g., carbon intensity) versus total-factor approaches integrating inputs, desired outputs, and undesirable outputs. DEA-based methods (CCR, BCC, SBM, DDF/NDDF) are favored over SFA for handling undesirable outputs and avoiding distributional assumptions. The global NDDF approach enables intertemporal comparability. Prior studies reveal regional and sectoral heterogeneity in carbon and energy efficiencies, but evidence at manufacturing subsector level remains limited. Second, determinants of MCEE: internal factors (notably R&D) and external factors (environmental regulation, openness/global value chain position), along with energy management, agglomeration, allocation efficiency, and digitalization. On firm scale, literature shows both benefits (economies of scale, financing, and strategic capabilities that can foster green tech) and drawbacks (management complexity, potential increases in emissions), with limited exploration of nonlinear effects and interactions with policy context. Gaps include simultaneously addressing the dual objective (growth and efficiency), using comprehensive efficiency measures, and examining how ER, R&D, and openness condition scale effects.

MCEE measurement: The study constructs a non-radial directional distance function (NDDF) under a global production technology to enable cross-year comparisons. Inputs are labor (average annual employees), capital (perpetual inventory method with depreciation following Chen, 2011), and energy (total energy consumption). Outputs include desired output (industrial value added) and undesirable output (CO2 emissions). Direction vector specifies contraction of inputs and emissions with expansion of desired output; equal weights are assigned across input, desirable, and undesirable components following prior work. CO2 emissions are computed via the energy consumption approach across eight fuels (raw coal, coke, crude oil, gasoline, kerosene, diesel, fuel oil, natural gas) using standard coal conversion, calorific values, oxidation rates, and emission factors. Data: Panel of 29 Chinese manufacturing subsectors (excluding three due to data gaps) from 2012–2021. Sources include China Statistical Yearbooks (industrial, energy, environmental, science and technology) and National Bureau of Statistics; variables are log-transformed. Variables: Dependent variable is MCEE (from NDDF). Core explanatory variable corporate scale (CS) is industrial value added per corporate unit. Threshold variables: environmental regulation (wastewater and waste gas treatment operating costs over main operating revenue), R&D investment (R&D funding over industrial value added), and trade openness (foreign plus Hong Kong/Macau/Taiwan investment over net fixed assets). Controls include energy endowment (energy consumption per labor), product structure (new product sales over main revenue), and market competition (non-state-owned capital over paid-in capital). Econometric strategy: - Pre-estimation tests: Panel unit root tests (LLC, IPS, Fisher ADF/PP) and Kao panel cointegration test to establish stationarity and long-run equilibrium. - Baseline models: OLS and fixed effects; dynamic System GMM with lagged dependent variable to address endogeneity (Arellano-Bond style instruments; Blundell-Bond system). Diagnostics include AR(1), AR(2), and Sargan tests. - Heterogeneity: Method of moments quantile regression (MMQR; Machado and Santos Silva, 2019) to capture location-scale effects and fixed effects across MCEE quantiles. - Nonlinearity: Panel threshold model (Hansen framework) with ER, RD, and OPEN as threshold variables; bootstrap tests for threshold significance; estimation of single-threshold models based on identified cutoffs and LR-based confidence intervals. Robustness: Difference GMM, alternative MCEE measure (Super-SBM), alternative proxy for firm scale (number of firms to average employees), and shorter sample period (2013–2020). Multicollinearity checked via VIF (<5).

- Data properties: First differences of variables reject unit roots across LLC, IPS, Fisher ADF/PP tests; Kao test indicates cointegration, supporting long-run relationships. - Baseline effects: System GMM shows lagged MCEE significant; corporate scale (InCS) coefficient = 0.0250 (1% level), implying a 1% increase in scale raises MCEE by about 2.5%. Diagnostics: AR(1) p < 0.1, AR(2) p > 0.1, Sargan p ≈ 0.225, indicating valid instruments and no second-order serial correlation. - Robustness: Across four checks—difference GMM, alternative MCEE measure (Super-SBM), alternative scale proxy, shortened sample—the CS coefficient remains positive and significant, with consistent AR/Sargan diagnostics. - Quantile heterogeneity (MMQR): The positive effect of corporate scale persists across MCEE quantiles, with coefficients declining from 0.1592 at Q0.1 to 0.1061 at Q0.9, indicating stronger scale benefits at lower-efficiency levels. Energy endowment and product structure are insignificant across quantiles; market competition becomes positive and significant from Q0.3 upward with increasing magnitude. - Threshold effects: Single thresholds are significant for ER, RD, and OPEN via bootstrap F-tests. Estimated thresholds: ln(ER) = -1.1811 (95% CI: -1.2222 to -1.1778), ln(RD) = 0.7827 (0.7435 to 0.7871), ln(OPEN) = 0.1075 (0.1054 to 0.1112). Conditional CS effects: • ER threshold: below threshold, CS coefficient 0.1186; above threshold, 0.2268 (both 1% level), indicating stricter ER amplifies the scale effect on MCEE. • RD threshold: below threshold, 0.0721; above threshold, 0.1386 (1% level), showing greater R&D strengthens the scale–MCEE link. • OPEN threshold: below threshold, 0.2374; above threshold, 0.1140 (1% level), suggesting deeper openness attenuates the positive scale effect, potentially via relocation or growth of pollution-intensive activities. - Descriptive/diagnostics: VIFs < 5; correlations show InCS and InPS positively associated with MCEE; InEE, InMC, InER, InRD, and InOPEN negatively correlated in raw correlations, though regression and quantile results refine these relationships.

The findings directly address the research questions by demonstrating that larger corporate scale significantly improves MCEE in China’s manufacturing subsectors, even after controlling for endogeneity and alternative specifications. Mechanistically, firm expansion likely enhances access to capital, technology, skilled labor, and managerial systems, enabling energy-saving process upgrades and better resource allocation, thus improving carbon efficiency. The MMQR results reveal that the scale effect is strongest among low-efficiency sectors, implying substantial gains from scale-driven modernization where production processes are less optimized. Conversely, at high MCEE quantiles, marginal scale benefits diminish, underscoring the importance of managerial sophistication and targeted innovation beyond sheer expansion. Threshold analyses show policy and market contexts condition the scale–MCEE link: stricter environmental regulation and higher R&D intensity magnify the gains from scale, consistent with induced innovation and compliance-driven efficiency improvements. In contrast, deeper trade openness weakens the positive impact, possibly due to increased pollution-intensive activity tied to foreign capital inflows or value-chain positioning. These nuanced interactions suggest that scale expansion must be complemented by robust environmental governance and sustained innovation investment to maximize efficiency gains, while openness policies should be steered toward cleaner sectors and technologies.

The study constructs a comprehensive MCEE measure using a global NDDF framework and empirically shows that corporate scale significantly and robustly enhances MCEE in China’s manufacturing subsectors (2012–2021). The positive effect exhibits heterogeneity across efficiency levels and depends nonlinearly on environmental regulation, R&D investment, and trade openness: stronger regulation and higher R&D amplify the benefits of scale, while greater openness dampens them. Contributions include: integrating firm scale and MCEE within a sector-specific efficiency framework; employing dynamic GMM, MMQR, and panel threshold models to capture endogeneity, distributional heterogeneity, and nonlinear policy interactions; and providing actionable thresholds for policy design. Future research should extend analysis to other emerging economies, explore additional mediating/moderating mechanisms (e.g., resource misallocation, ownership structure), and consider firm-level microdata to refine causal channels.

The analysis focuses on China’s manufacturing subsectors, which may limit generalizability to other countries and industries. While key contextual factors (environmental regulation, R&D, trade openness) are modeled as thresholds, other potential influences—such as resource misallocation, ownership structure, and more granular trade/value-chain positioning—are not explicitly incorporated. Measurement relies on sector-level aggregates and constructed proxies (e.g., scale, regulation), and CO2 estimations from energy use may omit process emissions outside the covered fuels. Future work with firm-level data and broader contextual variables could provide deeper causal insights and external validity.